用于超分辨率重建的同构过完备字典学习方法

J4 ›› 2014, Vol. 36 ›› Issue (08): 1441-1446.

用于超分辨率重建的同构过完备字典学习方法

谢宝陵1，徐国明1,2

(1.陆军军官学院基础部，安徽合肥 230031；2.合肥工业大学计算机与信息学院，安徽合肥 230009)

收稿日期:2012-10-31 修回日期:2013-02-25 出版日期:2014-08-25 发布日期:2014-08-25
基金资助:
安徽省自然科学基金资助项目(1208085QF115)

An isomorphic overcomplete dictionary learning
method for superresolution reconstruction

XIE Baoling1,XU Guoming1,2

(1.Department of Basic Sciences,Army Officer Academy,PLA,Hefei 230031;2.School of Computer and Information,Hefei University of Technology,Hefei 230009,China)

Received:2012-10-31 Revised:2013-02-25 Online:2014-08-25 Published:2014-08-25

摘要/Abstract

摘要：

构造合适的过完备字典是基于稀疏表示的超分辨率重建中的关键问题之一。在最大似然估计准则下，建立基于混合高斯的同构过完备字典学习模型。模型采用加权的l2范数来刻画分解残差，由分解残差设计权值矩阵，并且将同构的双字典学习问题转化为单字典的学习。采用稀疏编码和字典更新的交替迭代策略完成目标函数的求解，由内点法进行稀疏编码，采用拉格朗日对偶法完成字典更新。最后将学习得到的字典用于超分辨率重建实验，并与其他方法进行比较。实验结果验证了该模型和算法的有效性。

关键词: 超分辨率, 过完备字典, 混合高斯, 稀疏编码

Abstract:

Constructing an appropriate overcomplete dictionary is one of the key problems of superresolution based on sparse representation. In the maximum likelihood estimation principle, an isomorphic overcomplete dictionary learning model based on mixture Gaussian is proposed. Firstly, the sparse coding residual of the model is described by the weight l2norm and the weight matrix is designed by the residual. Secondly,the isomorphic coupled dictionary learning problem is translated into the single dictionary learning problem. The dictionary is learned by the alternate and iterative strategy using sparse coding and dictionary updating. An interiorpoint method is used in sparse coding and Lagrange dual is used in dictionary updating. Finally, the learned dictionary is used in the superresolution experiment,and compared with other methods.The experimental results demonstrate the effectiveness of the proposed method.

Key words: super-resolution;overcomplete dictionary;mixture Gaussian;sparse coding

谢宝陵1，徐国明1,2. 用于超分辨率重建的同构过完备字典学习方法[J]. J4, 2014, 36(08): 1441-1446.

XIE Baoling1,XU Guoming1,2. An isomorphic overcomplete dictionary learning
method for superresolution reconstruction [J]. J4, 2014, 36(08): 1441-1446.

参考文献［7］

［1］	Yang J C, Wright J, Huang T, et al. Image superresolution as sparse representation of raw image patches［C］∥Proc of IEEE Computer Vision and Pattern Recognition(CVPR),2008:18.
［2］	Yang J C, Wright J, Huang T, et al. Image superresolution via sparse representation［J］. IEEE Transactions on Image Processing, 2010, 19(11):28612873.
［3］	Zeyde R, Elad M, Protter M. On single image scaleup using sparserepresentations［C］∥Proc of the 7th International Conference on Curves and Surfaces, 2010:122.
［4］	Aharon M, Elad M, Bruckstein A M. The KSVD:An algorithm for designing of overcomplete dictionaries for sparse representation［J］. IEEE Transactions on Signal Processing, 2006, 54(11):43114322.
［5］	Pu Jian, Zhang Junping. Superresolution through dictionary learning and sparse representation［J］. PR&AI, 2010, 23(3):335340. (in Chinese)
［6］	Sun Yubao, Wei Zhihui, Xiao Liang, et al. Multimorphology sparsity regularized image superresolution［J］. Acta Electronica Sinica, 2010, 38(12):28982903. (in Chinese)
［7］	Lian Qiusheng,Zhang Wei.Image superresolution algorithms based on sparse representation of classified image patches［J］. Acta Electronica Sinica, 2012, 40(5):920925. (in Chinese)
［8］	Yang M, Zhang L, Yang J, et al. Robust sparse coding for face recognition［C］∥Proc of CVPR’11, 2011:625632.
［9］	Kim S J, Koh K, Lustig M, et al. An interiorpoint method for largescale l1regularized least squares［J］. IEEE Journal on Selected Topics in Signal Processing, 2007, 1(4):606617.
［10］	Lee H, Battle A, Raina R, et al. Efficient sparse coding algorithms［C］∥Proc of 2006 Conference on Advanced in Neural Information Processing Systems, 2007:801808.
［11］	Yang Jingyu, Peng Yigang, Xu Wenli, et al. Ways to sparse representation:An overview［J］. Science in China Series F:Information Science, 2009, 52(4):695703.
［12］	Rubinstein R, Zibulevsky A M, Elad M. Dictionaries for sparse representation modeling［J］. IEEE Transactions on Signal Process, 2010, 98(6):10451057.
［13］	Wright J, Ma Y. Dense error correction via l1 minimization［J］. IEEE Transactions on Information Theory, 2010, 56(7):35403560.
［14］	Boyd S,Vandenberghe L. Convex optimization［M］. New York:Cambridge University Press, 2009.
	附中文参考文献：
［5］	浦剑，张军平. 基于词典学习和稀疏表示的超分辨率方法［J］. 模式识别与人工智能, 2010, 23(3):335340.
［6］	孙玉宝，韦志辉，肖亮，等.多形态稀疏性正则化的图像超分辨率算法［J］.电子学报, 2010,38(12):28982903.
［7］	练秋生，张伟. 基于图像块分类稀疏表示的超分辨率重构算法［J］.电子学报, 2012,40(5):920925.

[1]	王万军, 丁新涛, 刘朝, 章智强. 基于多尺度联合网络的图像超分辨率重建方法[J]. 计算机工程与科学, 2023, 45(04): 683-690.
[2]	李彬, 喻夏琼, 王平, 傅瑞罡, 张虹. 基于深度学习的单幅图像超分辨率重建综述[J]. 计算机工程与科学, 2021, 43(01): 112-124.
[3]	刘艳萍1，崔彤1，周长兵2，李小翠2，刘甜1. 基于改进混合高斯模型和图形句柄的异常车辆检测[J]. 计算机工程与科学, 2020, 42(02): 266-272.
[4]	孙一飞，武继刚，张欣鹏. 面向眼底图像小目标检测的无监督学习方法[J]. 计算机工程与科学, 2019, 41(11): 2000-2006.
[5]	袁奕珊，陈姝. 基于稀疏表示的可变形部件模型目标检测[J]. 计算机工程与科学, 2017, 39(05): 999-1004.
[6]	孙锐，张广海，丁文秀. 基于分层特征融合的行人分类[J]. 计算机工程与科学, 2016, 38(10): 2115-2120.
[7]	李博川1，丁轲2. 结合阴影抑制的混合高斯模型改进算法[J]. J4, 2016, 38(03): 556-561.
[8]	叶立仁，何盛鸿，赵连超. 复杂环境下的遗留物检测算法[J]. J4, 2015, 37(05): 986-992.
[9]	黄玉，殷苌茗，周书仁. 基于YCbCr的自适应混合高斯模型背景建模[J]. J4, 2015, 37(01): 152-156.
[10]	徐蔚鸿，严金果. 基于改进的背景差分的运动目标实时检测算法[J]. J4, 2014, 36(07): 1352-1356.
[11]	晁永国. 一种改进的非负稀疏编码图像编码方案[J]. J4, 2010, 32(10): 66-68.
[12]	晁永国, 戴芳, 韩舒然, 何静. 改进的非负稀疏编码图像基学习算法[J]. J4, 2010, 32(1): 77-79.
[13]	单勇[1,2] 王润生[2] 杨凡[1]. 基于多特征背景模型的运动目标检测算法[J]. J4, 2007, 29(8): 40-42.
[14]	刘英锋王程王润生. 图像序列超分辨率处理中充分帧数的估计[J]. J4, 2007, 29(1): 59-61.
[15]	方景帅窦文华钱悦. 一种基于DSP的运动目标检测与判别算法[J]. J4, 2006, 28(3): 49-52.